TI 2013-115/VIII Tinbergen Institute Discussion Paper

Dynamics in Aceh and North Sumatera after the Twin Disasters

Aloysius Gunadi Brataa,b Henri L.F. de Groota,c,d Piet Rietvelda,c

a Faculty of Economics and Business Administration, VU University Amsterdam, The Netherlands; b Atma Jaya University, Department of Economics, Yogyakarta, ; c Tinbergen Institute, Amsterdam, The Netherlands; d Ecorys NEI, Rotterdam, The Netherlands.

Tinbergen Institute is the graduate school and research institute in economics of Erasmus University Rotterdam, the University of Amsterdam and VU University Amsterdam.

More TI discussion papers can be downloaded at http://www.tinbergen.nl

Tinbergen Institute has two locations:

Tinbergen Institute Amsterdam Gustav Mahlerplein 117 1082 MS Amsterdam The Netherlands Tel.: +31(0)20 525 1600

Tinbergen Institute Rotterdam Burg. Oudlaan 50 3062 PA Rotterdam The Netherlands Tel.: +31(0)10 408 8900 Fax: +31(0)10 408 9031

Duisenberg school of finance is a collaboration of the Dutch financial sector and universities, with the ambition to support innovative research and offer top quality academic education in core areas of finance.

DSF research papers can be downloaded at: http://www.dsf.nl/

Duisenberg school of finance Gustav Mahlerplein 117 1082 MS Amsterdam The Netherlands Tel.: +31(0)20 525 8579

Dynamics in Aceh and North Sumatera after the Twin Disasters

An investigation into the relevance of the locational fundamental theory

Aloysius Gunadi Brataa,b, Henri L.F. de Groota,c,d and Piet Rietvelda,c

a VU University Amsterdam, Department of Spatial Economics, The Netherlands b Atma Jaya Yogyakarta University, Department of Economics, Indonesia c Tinbergen Institute, Amsterdam, The Netherlands d Ecorys NEI, Rotterdam, The Netherlands

Abstract This paper analyzes the impact of the Indian Ocean Tsunami in 2004 and the Nias earthquake in 2005 on population dynamic across regions in Aceh and North Sumatera. We find no evidence that the disasters affected the regularity of size-distribution of the regions. The estimates of a population growth model yield clear evidence that the disasters had only a temporary impact. This study provides indicative evidence for the relevance of the locational fundamental theory and rejects the random growth explanation.

Keywords : natural disaster; population growth; rank-size distribution JEL code : R11, R12, Q54

“A city is hard to kill, in part because of its strategic geographical location, its concentrated, persisting stock of physical capital, and even more because of the memories, motives, and skill of its inhabitants.” (Kevin Lynch, 1972: 3-4, cited by Vale and Campanella 2005: 347)

1. Introduction There are several stylized facts with regard to the growth of cities across countries. Cities in many countries are growing according to a regular pattern, both in terms of their size (city population) and their rank in a system of cities. The regularity over decades or centuries means that even large shocks did not affect the distribution of cities permanently. This regularity became a fascinating ‘law’ for economists to evaluate and to explain theoretically (see, for instance, Gabaix and Ioannides, 2004; Krugman, 1996; Nitsch, 2005). Another dimension concerns the location of cities. Vale and Campanella (2005) conclude that one of the city resilience axioms is the power of place which contributes to city resilience after it was destructed by a major disaster. They derived this from studies on the resilience of cities in different time periods, places and cultures. Although this conclusion is based on a limited number of studies on individual cities that have been destroyed by large shocks, it provides evidence that cities which have locational advantages will rebound and subsequently keep their high rank in the distribution of cities. In the same spirit, Davis and Weinstein (2002), Brakman et al. (2004) and Bosker et al. (2008) used the bombing of Japan and German during the second World War (WW-II) as “quasi natural experiments” and found that the impact of shocks on the city-size distribution is only temporary and supports the locational fundamental theory. Investigating another post WW-II “natural experiment” – the bombing of Vietnam by the US – Miguel and Roland (2011) did not find strong evidence of permanent effects from this bombing on local poverty rates, consumption levels, infrastructure, literacy, or population density. They used a different approach, but the main conclusion is parallel to Japan and German cases, which is that the bombing affected Vietnam only temporary. Basically, locational fundamentals theory states that natural advantage or the physical landscape (first nature advantages such as access to the sea, river, or availability of natural

1 resources), as initial conditions, determines the existence as well as the growth of cities (see, for instance, Fujita and Mori, 1996; Krugman, 1996). These permanent features give benefits to the locations as an excellent site for economic activity (Davis and Weinstein, 2002) and as such have affected the formation and evolution of the size of a location. It is oftentimes argued that in normal situations, the importance of such initial conditions should become smaller, but their impacts may still persist, for instance through cumulative process by which first nature advantages create second nature advantages. Against this background, it is interesting to see how the growth and the position of a city in the rank size distribution is affected by extreme events: do the initial locational advantages play an important role in reverting the growth of the city after the shock? Following Davis and Weinstein (2002), the main aim of this paper is to investigate the relevance of the locational fundamental theory in a very specific situation, viz. the cross- regions in Northern Sumatera (Aceh and North Sumatera province) after the Indian Ocean tsunami and the Nias earthquake. Since most empirical studies on urban growth focus on the developed countries, taking Northern Sumatera as a “natural laboratory” will contribute to recent empirical studies on developing countries such as for instance China (Anderson and Ge, 2005; Chen et al., 2010), Malaysia (Soo, 2007), or on growth of sub-national (state) population in India, China and Brazil (Soo, 2010). Another important issue in studying this size distribution to which we aim to contribute is the appropriate urban-regional unit of observation. In their recent papers, Giesen and Suedekum (2011) and González-Val and Sanso-Navarro (2010) argue that Zipf’s law can also hold within regions of a country. Applying three different concepts of regions in Germany (random regions, the German Federal States, and a spatial club of cities), Giesen and Suedekum show that city size distributions at national and regional levels tend to follow a power law. A similar message on this issue also emerged when regions in the USA were clustered as “Megapolitan Areas” (see Berry and Okulicz-Kozaryn, 2011). These studies argue in favour of using data for single regions. This may be particularly relevant for regions such as Aceh and Nias in large developing countries such Indonesia. Based on our empirical work, we conclude there are no significant changes of relative positions of regions in Aceh and North Sumatera, despite the fact that they have been hit by big disasters. Post- and pre-disaster ranks are highly correlated as well as population size.

2

Furthermore, the twin disasters did not affect the regularity of the size distribution of regions in Aceh and North Sumatera. The estimated population growth model yields clear evidence that the twin disasters had only a temporary impact. Overall, this study provides indicative evidence that the locational fundamental theory holds. The rest of this paper is structured as follows. The next section provides some relevant background. In Section 3, we discuss our framework to investigate the impact of the twin disasters on population dynamics. We proceed with results and discussion in Section 4. Secction 5 concludes.

2. Setting the Scene The Indian Ocean tsunami occurred after a massive earthquake with 9.3 magnitudes (on the Richter scale) of the west coast of on December 26, 2004. The Nias earthquake occurred a couple of months later on March 28, 2005 (see Figure 1). There is no doubt that these natural disasters meet the basic criteria for ‘natural experiments’ in that they are exogenous, large, variable, and purely temporary (cf. Davis and Weinstein, 2008). More than 300,000 people were killed by the giant tsunami which also resulted in huge physical destruction. Meanwhile, the Nias earthquake with 8.7 magnitudes on the Richter scale killed more than 900 people, and strongly affected the Nias island. In response to these natural disasters, the Government of Indonesia (GoI) established the Agency for the Rehabilitation and Reconstruction of Aceh and Nias.1 The operation of this agency is regarded as one of the largest humanitarian programmes in history with nearly 500 participating actors (see Takahashi et al., 2007; Masyarafah and McKeon, 2008).

1 This agency, which is locally known as BRR (Badan Rehabilitasi dan Rekonstruksi), was established in 2005 and was operational for four years, based in with a regional office in Nias and a representative office in . It coordinated and jointly implemented a community-driven recovery program for Aceh and Nias. See http://kc.monevacehnias.bappenas.go.id/Modules/Home-Accordion/about-brr.html.

3

Figure 1. Epicentres of the twin disasters

Source: Nazara and Resosudarmo (2007), with permission. We modified the map by adding the yellow and blue boxes.

4

Table 1. The impact of the twin disasters in Aceh and Nias Internally Displaced House Conflict No Region Deaths and missing People Damaged Intensity per 1000 per 1000 Index total people* total people* % 1 Aceh Barat 14854 86.00 67817 392.63 45 1.49 2 Aceh Barat Daya 9 0.08 3480 30.13 13 0.56 3 Aceh Besar 47784 161.30 97466 329.00 32 0.74 4 Aceh Jaya 19804 248.39 40422 507.00 61 1.75 5 Aceh Selatan 9 0.05 19049 97.87 10 1.56 6 Aceh Singkil 50 0.37 30967 226.84 6 0.80 7 Aceh Tamiang 4 0.02 3396 14.75 1 0.69 8 Aceh Tengah 209 0.78 6107 22.65 2 1.90 9 Aceh Tenggara 172 1.08 809 5.06 2 0.87 10 Aceh Timur 224 0.70 14411 45.30 1 3.63 11 Aceh Utara 2726 5.36 32246 63.39 8 1.64 12 Bireuen 1261 3.54 42143 118.15 9 1.04 13 Gayo Lues 3 0.04 0 0.00 1 1.50 14 Nagan Raya 1338 10.00 17040 127.29 9 2.11 15 Pidie 6109 12.32 81532 164.37 8 1.65 16 Simeulue 45 0.65 42751 621.91 45 0.22 17 Banda Aceh 77804 387.37 50970 253.77 59 0.00 18 Langsa 0 0.00 3489 26.79 0 0.00 19 Lhokseumawe 404 2.51 7577 47.06 6 1.35 20 Sabang 14 0.53 3712 139.82 6 0.00 21 Nias 967 1.37 70000 99.25 79 0.00 Mean of sample (threshold value) 4456 23.65 16292 85.46 10 0.60

Sources of data: see Appendix. * Population data are an average of the population in 2003 and 2005. Regions in North Sumatera that were not affected by the twin disaster were excluded from this table.

Table 1 provides figures of the impact of the disasters on the population of Aceh and Nias. We do not provide data for other regions in North Sumatera since they were not affected, but we include them as control regions in our empirical analysis. It is important to note that regions in this table have been aggregated using 2003 as the reference point in order to avoid inconsistencies in the data caused by changes in administrative boundaries related to the decentralization. One of the impacts of decentralization in Indonesia is proliferation of

5 administrative regions across the country (cf. Fitrani et al. 2005). As a result, the number of regions in Aceh and North Sumatera has been increasing. We selected 2003 as the reference point by considering that the number of regions – especially in Aceh before 2003 – is quite small. Regencies or cities created since 2003 have been merged with the parent regions in our dataset.2 Using this procedure, Aceh in our dataset consists of 16 regencies and four cities while North Sumatera has 13 regencies and six cities. Thus, there are 29 regencies and 10 cities in our sample. Overall the table reveals the magnitude of the disasters in terms of the impact on human beings as well as on houses. For the purpose of our study, we classified the regions into two groups: (1) highly affected regions and (2) non- or limitedly affected regions. We adopted a method that has been used by Cavallo et al. (2010) to identify large international disasters during the period 1970–2008. The threshold values for classifying the regions are the means of the density of disaster impacts (number of deaths and missing per 1000 people, number of Internally Displaced People per 1000 people and percentage of houses damaged). A region is classified as a highly affected region if at least one of these three indicators is greater than the relevant threshold. Based on this procedure, 13 regions, mostly in the western and southern coastal areas of Aceh and the Nias island of North Sumatera, were categorized as highly affected regions (bold numbers in Table 1 and yellow boxes in Figure 1). Regarding population dynamics, it is known that compared to other provinces in Indonesia, the population growth rate in Aceh has fallen drastically especially between 2000 and 2005 (BPS-Statistics et al., 2010, pp. 16–17). The three key factors explaining this drastic fall are (i) higher mortality due to the tsunami, earthquakes and civil war, (ii) a falling birth rate, and (iii) outmigration. Besides increased mortality, the military conflict also contributed to a falling birth rate. Both the conflict and the weakening of the economy accelerated outmigration from Aceh. The peace deal and massive recovery programs after the 2004 tsunami led to an increase in population growth in Aceh. The Population Census 2010 confirms the role of migration in explaining Aceh’s population increase. It shows that recent migration in Aceh is 0.6 percent, while in 2000 it was equal to –9.5 percent (BPS-Statistics, 2010).

2 For instance, Aceh Tengah in Table 1 includes Bener Meriah (see Figure 1), a new region that was founded in 2003.

6

For Aceh in particular, the tsunami also opened a ‘window of opportunity’ to end the thirty years of military conflict between the Free Aceh Movement and the Government of Indonesia (see, for instance, Nazara and Resosudarmo, 2007; Billon and Waizenegger, 2007).3 Furthermore, the Nias Island is expected to have an opportunity to improve its local development quality and to reduce poverty through the so called ‘build back better’ approach in reconstruction and rehabilitation programs.4 This opportunity is important since although the number of fatalities in Nias is smaller than in Aceh, the physical damages caused by the earthquake are evenly spread across a large part of the Nias Island (BRR Aceh-Nias Perwakilan Nias, 2008).

3. A framework to study the impact of the twin disasters on population dynamics As already stated in the introduction, we follow an approach developed by Davis and Weinstein (2002) to empirically investigate the relevance of the locational fundamental theory in Northern Sumatera after the twin disasters. There are three theoretical strands that have been used in studying the impact of shocks on the distribution of cities, viz. increasing returns to scale, random growth, and locational fundamentals. According to Bosker et al. (2008), all three theoretical strands predict a stable city size distribution in equilibrium, but they yield a quite different reaction to shocks. With increasing returns, a large shock has the potential to (radically) change the city size distribution. Random growth theory predicts a permanent effect of shocks on city sizes and on the relative position of cities within the distribution. Finally, the locational fundamental theory argues that a large shock has a temporary effect only on relative city size as far as the shock does not change the underlying locational fundamentals themselves (i.e., the first nature geography). Regarding the findings in Davis and Weinstein (2002) that the Japanese case supports the third theory, Polèse and Dennis-Jacob (2010) find that actually the characteristics of a

3 On December 28, 2006, two days after the tsunami, Aceh was opened to outsiders. To be more precise, the agreement of peace (the Helsinki Memorandum of Understanding) was officially signed on August 25, 2005, by Hamid Awaluddin (chief Indonesian negotiator) and Malik Mahmud (leader of the Free Aceh Movement) after a series of negotiations. These negotiations were moderated by former Finnish President Martti Ahtisaari. In 2008, Ahtisaari received the Nobel Peace Prize for his significant efforts over more than three decades to resolve international conflicts. 4 ‘Build back better’ or ‘building back better’ was the tag line or mantra which many groups gave to the post- disaster reconstruction. Its ambition is to do more than just restore Aceh and Nias to their previous conditions. See, for instance, Kennedy et al. (2008).

7 country may determine which theory is most applicable. For instance, there is a difference between countries with mature urban systems and developing nations in the probability that the top city will be pushed aside by an upstart. This probability is rather small in mature urban systems, but it is more likely to materialize in developing nations. They also noted that ‘fundamental advantages’ are in part a function of technology and preferences: the manifestation of ‘fundamental’ was different in the past (for instance, ports were fundamental for transoceanic travel) than it is these days (for instance, part of travel through ports having been replaced by air travel). In short, based on their study on the dynamic changes of the top cities in national urban hierarchies, they concluded that political events and technological change can disturb the ‘fundamental’ advantages of the first big cities. When the locational fundamental theory applies in the context of Northern Sumatera, we should not find permanent changes in the rank of affected regions nor of non-affected regions after the twin disasters. However, it is reasonable to expect that there are rank changes in the aftermath of disasters and that it takes years before they are back to their pre-disaster rank. In checking the expectations, it is useful to compare the regions’ rank after the disaster (viz. in 2005 and 2010) with the 2003 rank along with the computation of simple correlations between rank and size in pre- and post-disaster periods. This can then be followed by relating rank and size of population by using the Zipf equation as the standard procedure. Regarding this Zipf equation, Gabaix and Ibragimov (2009) noted that for small samples, the estimated power law coefficient in the classic OLS Zipf regression is biased and inefficient. The proposed procedure to solve this problem is to shift the rank by 0.5. Then, the OLS model for the Zipf equation to be estimated is:

log(Ri – 0.5) = α – β·log(Pi) + εi , (1)

where R is the rank of region i in the system of regions (based on population), Pi is the population of region i, and β is the power component that is our main coefficient of interest. Following Gabaix and Ibragimov (2009), the standard error of the estimated power component (b) is computed as [√(2/N)]·b.

8

3.1. The spatial aspect of disaster on population growth The description of the change of rank of regions will provide a useful first indication whether the tsunami and earthquake had permanent impacts on population dynamics in these regions. The main concern here is to find out the impact of disasters on population growth by adopting an empirical model that has been used in Davis and Weinstein (2002) and Brakman et al. (2004). The basic model in exploring whether a disaster has a permanent or temporary impact on the population growth is as follows:

· , (2)

where is the average annual population growth between 2005 and 2010 (in percentage points) and is the average annual population growth between 2003 and 2005 (in percentage points). If ε = 0, population growth follows a random walk; if ε = –1, the disaster had no effect at all; and finally, if –1 < ε < 0, then the disaster only had a temporary impact. The population growth rate on the right-hand side of equation (2) will be instrumented by using disaster variables. Two disaster variables will be employed as follows:

= ζ + η·DM + ν·HD, (3)

where is the predicted value of population growth between 2003 and 2005 (in percentage points), DM is number of deaths and missing people per 1000 people and HD is the percentage of houses damaged. An additional element of our approach concerns the spatial process of population growth after disasters. As shown in Table 1, the twin disasters have killed thousands of people and have resulted in many Internally Displaced People (IDP) as well as destroyed houses. The direct impact of these disasters on the affected regions was that their population suddenly dropped. In responding to large disasters, individuals or families in affected regions may decide to move to the nearest regions that were not affected by disasters. This is a coping mechanism with the disasters by searching employment in non affected regions and may also be a response to the fear for subsequent disasters in the near future (Paul, 2005). In the case of

9

Aceh, people moved from the coastal areas that were most affected by the tsunami to their villages of origin (locally known as gampöng) – the important last resort for Acehnese in more peripheral areas (Mahdi, 2007).5 If this would be a temporary move, then one could say that disasters have a temporary impact only on population, which supports the locational fundamental theory. This raises an empirical question on spatial effects in the form of population spill-overs from affected to neighbouring regions. In measuring neighbourhood impacts on population growth in the context of the disasters, we use a dummy variable (SB_HAR) to classify whether a region shares an administrative border with a highly affected region or not (the dummy equals 1 for regions that share a border with highly affected regions, and equals 0 otherwise). We expect that regions that are neighbours of highly affected regions will receive positive spill-overs from their highly affected neighbours. There are two possible explanations for such spill-overs. The first is an out-migration from their affected neighbours. The second one is a spatial ‘benefit’ spill-over that may result from relatively high population growth in the highly affected region in the post-disaster years. However, as indicated by the results of the 2000 Census, Aceh experienced in-migration in 2010 (BPS-Statistics 2010), so the most reasonable explanation is not the first, but the second one.6 Regarding the spill-overs, it is important to note that most highly affected regions share common borders. Therefore, we split regions in two types of groups. The first is a group of highly affected regions that share borders with other highly affected regions, denoted as HAR_SB_HAR (1 for highly affected regions that share borders with other highly affected region, and 0 otherwise). The second is a group of non- and low-affected regions that share border with highly affected regions, denoted as NHAR_SB_HAR (1 for non-low affected regions that share border with highly affected region, and 0 otherwise). The expectation is that spatial affects captured by the first variable will be stronger than those captured by the second

5 Gampöng represents not only an administrative territory, but also the important traditional or informal networks for Acehnese (see also Gayatri, 2009, for recent dynamics of gampöng). 6 Regarding migration flows from the Tsunami-affected areas and others from conflict-affected areas suffering poor economic conditions in Aceh to Malaysia, Naik et al. (2007) argued that there is no evidence available to show whether numbers increased post-Tsunami. Furthermore, they mentioned that the peace agreement in Helsinki had the potential to result in the return of large numbers of transmigrants and Acehnese.

10 variable. We then arrive at the two following models which are refined versions of equation (2):

= δ + ε · + ρ·SB_HAR (4)

= δ + ε · + ρ1·HAR_SB_HAR + ρ2·NHAR_SB_HAR (5)

The spatial impact of a highly affected region on population growth of their neighbours is captured by ρ. If the neighbour is also a highly affected region, the impact will be shown by ρ1, while ρ2 captures the impact for non- and low affected regions.

3.2. Other relevant variables In this section we discuss other relevant variables to explain population growth. There are three additional factors that will be considered in exploring the impact of disasters on population dynamics: (i) the recovery program, (ii) the peace deal, and (iii) institutional factors. For the recovery program, it is clear that the rehabilitation and reconstruction fund is a potentially important determinant of the recovery process. Davis and Weinstein (2002) show that government reconstruction expenses positively influenced the recovery of Japanese cities after the WW II bombing. However, they also note that this impact is quite small due to the focus of reconstruction policies on rural areas. Brakman et al. (2004) found similar results for West Germany since there was not a guarantee that the most destroyed cities during the war received more funds and since the government tended to support large cities. These examples indicate that rehabilitation support does not necessarily have a large impact on post-shock growth. In the case of Aceh and Nias after the disasters, Masyarafah and McKeon (2008) noted that in 2005 reconstruction funds were disproportionally allocated across the affected areas. As measured by geographical gaps in the allocation of funds (ratio of financing to needs), two administrative units that received relatively many funds were Banda Aceh and Aceh Besar.7

7 Details of the geographical gaps can be found in Masyarafah and McKeon (2008, Map 1, p. 33).

11

As shown in Table 1, both regions are categorized as highly affected regions, but it should be noted that Banda Aceh is the capital of the province of Aceh while Aceh Besar is the nearest neighbour of Banda Aceh. With better accessibility and as the home base of the rehabilitation agency, Banda Aceh and its direct neighbour were at a position to receive more funds than other hard-hit areas such as Aceh Jaya. Besides, it seems that rehabilitation and reconstruction programs for Aceh and Nias were started with a domination of a sectoral approach rather than a regional approach. Later in 2006, the rehabilitation agency applied ‘regionalisation’ where the authority was dispersed into several areas.8 We clarify the impact of recovery programs by adding a variable measuring the disbursed funds (in million USD), Disbursed_RAN, to the basic model. In line with previous studies, we do not expect that this variable has a large impact on population growth. During the military conflict, Aceh depended heavily on extractive industries in a few locations and the result was an uneven distribution of economic activities in Aceh. These extractive industries were oil and gas and their contribution to Aceh’s Gross Regional Product (GRP) had decreased from more than 50 percent in 2003 to around 20 percent in 2008 as a result of depletion of gas resources (World Bank, 2009). As extractive industries, the oil and gas sector in this east coast area of Aceh (Aceh Timur, Aceh Utara, Aceh Tamiang and Lhokseumawe) only have limited linkages with the rest of the economy of the region. The twin disasters are expected to reduce this inequality through the ‘positive’ impact of disasters in providing a chance for Aceh to become a more open province following the end of the military conflict that mainly affected the rural areas. As argued in World Bank (2009), many Acehnese were enabled to return to daily activities and resume work in post-conflict years. The report also mentioned that the low growth of non-oil and gas sector partly reflects the impact of conflict. It is important to note that the sequence in Aceh is contrary to other countries where big earthquakes have been followed by political revolution, for instance the Islamic Iranian

8 The locations of the regional offices of this agency are: (1) Lhokseumawe office servicing Pidie, Bireuen, Lhokseumawe and Aceh Utara; (2) Langsa office servicing Aceh Timur, Langsa, and Aceh Tamiang; (3) Kutacane office servicing Aceh Tengah, Aceh Tenggara, Gayo Lues and Bener Meriah; (4) Singkil office servicing Aceh Selatan and Singkil; (5) Meulaboh office servicing Aceh Barat, Nagan Raya and Aceh Barat Daya; (6) Sinabang office servicing Simeulue; (7) Calang office servicing Aceh Jaya; (8) Banda Aceh office servicing Aceh Besar, Banda Aceh and Sabang; (9) Nias office servicing Nias and Nias Selatan. See BRR Aceh- Nias, Multi Donor Fund, and UNDP (2006, p. 10).

12 revolution after the 1978 earthquake and the Sandinista revolution a few years after the earthquake in Managua Nicaragua (Cavallo et al., 2011). Cavallo et al. (2011) concluded that only large natural disasters followed by radical political revolution had a long-run negative effect on economic growth. Regarding this finding, one may hypothesize an opposite (positive) effect when a big natural disaster is followed by a peace deal. Therefore, in this study we expect that having a peace deal will benefit regions that suffered from conflict in the past, as measured by conflict intensity index, abbreviated as CII. It means that these regions will get a better opportunity to grow faster than others. However, since some of the high conflict regions were also strongly affected by disasters, it is important to check the cross effect between disaster impact and conflict index. This interplay variable is denoted as CII·HAR, where HAR is a binary variable that equals 1 if the region is highly affected (and 0 otherwise). Finally, regarding the role of institutional factors on the regional dynamics of population (see Polèse and Dennis-Jacob, 2010) a dummy variable for the provincial capital (D_Prov_capital) is introduced. In their cross-country study of the dynamics of the top cities in national urban hierarchies, Polèse and Dennis-Jacob found that changes in the rank at the very top are rare and that such changes are closely related to political events. In our sample, this variable is attributed to Banda Aceh – the capital of Aceh province, and – the capital of the province of North Sumatera.9 We expect that provincial capitals are at a better position to achieve high population growth.

4. Results and discussions

4.1. A description of regularity of size-distribution Figures 2, 3, and 4 show changes in the rank based on population from 2003 to 2005, from 2005 to 2010 and from 2003 to 2010.10 Figure 2 shows that there are four out of 13 regions that were strongly affected by the disasters as their rank dropped three levels. These are regions situated at the east coast of Aceh. Table 1 confirms that most fatalities of the disasters were residents of these regions. So these drops are most likely related to the sudden impact of

9 We provide details of variables used in this paper, as well as their sources, in Appendix 2 (Data Sources). 10 See Appendix 1 for abbreviations.

13 the disasters. Five highly affected regions survived at their pre-disaster rank. Meanwhile, only five out of 26 non- or low-affected regions dropped their rank. We subsequently investigate whether the highly affected regions are able to recover from the sudden impact of the twin disasters. Figure 3 compares the rank in 2005 with the rank in 2010. It shows that there is only one of the highly affected regions in the left quadrant whereas 8 regions survived at their 2005 rank. Furthermore, some severely affected regions in the left quadrant in Figure 2 moved to the right quadrant in Figure 3. Figure 4 shows a summary of these processes: only few of the highly affected regions could not regain their pre- disaster rank. These figures indicate that the impact of the disasters tends to be temporary only.

Figure 2. Change in rank between 2003 and 2005

14

Figure 3. Change in rank between 2005 and 2010

Figure 4. Change in rank between 2003 and 2010

15

Figure 5. Comparison of population size: pre- and post disasters

A comparison between pre- with post-disaster population size also confirms that most of the affected regions completely recovered from the impact of the twin disasters (Figure 5). Post- disaster population of five highly affected regions is still less than population size in 2003. In other words, more than a half of highly affected regions show a significant population growth. It took only five years for them to completely recover in terms of their population size. Therefore, it is not surprising that post- and pre disaster ranks are highly correlated as well as population size (Table 2).

Table 2. Correlation of rank of - and size of population 2003, 2005, 2010

Rank of 2003 population All Highly affected Non- and low affected (39) (13) (26) Rank of 2005 population 0.9929 0.9849 0.9966 Rank of 2010 population 0.9933 0.9882 0.9945 Size of 2003 population All Highly affected Non- and low affected (39) (13) (26) Size of 2005 population 0.9985 0.9915 0.9991 Size of 2010 population 0.9974 0.9929 0.9976

16

Overall, the findings indicate that the twin disasters as large shocks did not affect the regularity of the size-distribution of regions in Aceh and North Sumatera. Estimations of Zipf regressions for all regions – affected regions and non affected regions (Table 3) – show power components that are between 0.914 and 1.012 (or close to 1). These findings confirm that Zipf’s law also applies in Aceh and North Sumatera regions. In other words, Zipf’ law applies even in the group of regions that were highly affected by the giant tsunami and hazardous earthquake. So, this also suggests that the tsunami and earthquake did not affect population dynamics permanently.

Table 3. Zipf estimates 2003, 2005, 2010

Dependent variable: log(R – 0.5) Constant log(P) R-squared All regions (N =39) 2003 6.278*** (0.737) –0.943*** (0.214) 0.843 2005 6.297*** (0.667) –0.945*** (0.214) 0.861 2010 6.416*** (0.694) –0.962*** (0.218) 0.850 Affected regions (N =13) 2003 5.469*** (1.276) –0.914*** (0.358) 0.777 2005 5.771*** (1.144) –0.975*** (0.382) 0.829 2010 5.783*** (1.233) –0.968*** (0.380) 0.808 Non affected region (N=26) 2003 6.515*** (0.698) –1.001*** (0.278) 0.882 2005 6.568*** (0.684) –1.007*** (0.279) 0.889 2010 6.613*** (0.690) –1.012*** (0.281) 0.890 Notes: Robust standard errors in parentheses; *** Significant at 1%; ** Significant at 5%; * Significant at 10%. The standard error for logP equals [√(2/ N)]·b (cf. Gabaix and Ibragimov, 2009).

4.2. Estimating disaster impact on population growth Our descriptive analysis has revealed that the impact of the twin disasters on regions in Aceh and North Sumatera seems only temporary. In this section, we further investigate this empirically along the lines set out in Section 3. If the estimation results show evidence that the Indian Ocean tsunami and the Nias earthquake have only a temporary impact, then, following Davis and Weinstein (2002) one might reject the random growth theory and support the

17 locational fundamental theory. We will start with the basic model and it will be followed by the extended model.

4.2.1 Results of the basic model First, we estimate equation (3) to instrument population growth in the period 2003–2005. Table 4 shows that deaths and missing people (DM) significantly affected population growth between 2003 and 2005. However, coefficients of damaged houses are not statistically significant in the second specification although they have the expected negative sign. We therefore instrument past population using DM only.

Table 4. Estimation results (dependent variable is g0305)

Variable 1 2 Constant 4.742*** 4.957*** (1.498) (1.262) DM –0.098*** –0.092** (0.031) (0.040) HD –0.034 (0.152) R2 0.379 0.381 Notes: Robust standard errors in parentheses; *** Significant at 1%; ** Significant at 5%; * Significant at 10%.

Table 5 provides estimation results for the basic equation of population growth using the instrumented variable ( . These results consistently show a negative impact of ranging from –0.504 (specification 3) to –0.674 (specification 1). Since all coefficients are in between –1 and 0, we conclude that the twin disasters only had a temporary impact on regions in Aceh and North Sumatera. Within a short period of time of about five years, about 50 percent of the sudden impact of disasters on population growth has disappeared. This speed of recovery is faster than in Japan (15 years) and West Germany (more than 17 years) as estimated in Davis and Weinstein (2002) and Brakman et al. (2004), respectively. Acknowledging the short period covered by this study, the finding supports the locational fundamental theory that states that the impact of shocks on growth is only temporary.

18

Table 5. Estimation results (dependent variable is g0510)

Independent variable 1 2 3 Constant 8.567*** 5.726*** 5.449*** (0.930) (1.489) (1.503) *** *** *** –0.674 –0.562 –0.504 (0.112) (0.094) (0.090) SB_HAR 5.275*** (1.683) HAR_SB_HAR 6.837*** (1.818) NHAR_SB_HAR 3.783* (1.909) R2 0.447 0.559 0.576 Notes: Robust standard errors in parentheses; *** Significant at 1%; ** Significant at 5%; * Significant at 10%.

It is also important to underline that the results in Table 5 confirm that there is a spatial impact on the population growth. As shown in specification (2), the coefficient of SB_HAR is statistically significant indicating that there is a difference between population growth of regions that share a border with highly affected regions and population growth of other regions. It means that there is a ‘benefit’ of being situated next to an area that suffered from the disasters since the affected regions have a positive impact on their neighbours. By splitting SB_HAR into two groups (specification 3), the result confirms that the spatial effect among the affected regions is larger than the spatial effect between affected regions and non-affected regions. In other words, there is a reinforcing process among highly affected regions. In general, the results in this table suggest that population growth of regions neighbouring highly affected regions is relatively high.

4.2.2 Results of the extended model

Table 6 presents results for extended models. First, the results show that has a consistently negative impact on population growth in post-disaster years. Comparing their magnitude in the extended specifications (Table 6) to the magnitude in the basic specifications

(Table 5) it is clear that adding relevant variables does not disrupt the impact of . Therefore, in general, these results confirm that the locational fundamental theory holds in this

19 case since the negative coefficients indicate that the twin disasters only had temporary impact on regions. Meanwhile, the spatial variables (SB_HAR and its components) tend to show a decrease in their significance level, but the sign of their coefficients is robust. It means that basically the spatial spill-over still exists, especially among highly affected regions as shown by HAR_SB_HAR in specifications (4) and (5). In line with Davis and Weinstein (2002) and Brakman et al. (2004) for Japan and West Germany after the WW II bombing, respectively, reconstruction funds do not have a statistically significant impact on population growth. One possible reason for this finding is the specific regional distribution of recovery funds as has been noted in Masyarafah and McKeon (2008). Coefficients of the institutional factor that is represented by D_Prov_capital are statistically insignificant in all specifications, so there is no statistically significant difference in population growth between regions that serve as a provincial capital and other regions. As expected, CII which represents conflict experience in the past always has a positive and significant impact in all specifications. It indicates that the peace deal has an important role on growth of population in these regions. Regions that suffered from past conflict have benefited from the peace deal between the Government of Indonesia and the Free Aceh Movement. Furthermore, by adding the interaction variable between conflict and disaster (CII·HAR) to the model (specifications 3, 4 and 5), the coefficient of CII increases substantially from 3.8 to 4.4. Since the coefficient of the interaction variable is negative, it indicates that the benefit of the peace deal is lower in the regions that were also highly affected by disasters. In other words, the overall positive impact of peace deal is particularly high if there were no disasters.

20

Table 6. Estimation results, extended (dependent variable is g0510)

Variable 1 2 3 4 5 Constant 4.297** 4.340** 4.382** 4.280** 3.990*** (1.669) (1.711) (1.716) (1.708) (1.368) *** *** *** *** *** –0.510 –0.510 –0.600 –0.560 –0.524 (0.125) (0.132) (0.134) (0.139) (0.073) SB_HAR 2.530* 2.294 (1.448) (1.412) HAR_SB_HAR 3.207 3.890* 3.745* (1.939) (1.979) (2.038) NHAR_SB_HAR 2.048 1.957 2.049 (1.572) (1.691) (1.606) Disbursed_RAN 0.006 0.004 –0.002 –0.000 (0.007) (0.008) (0.008) (0.008) CII 3.816*** 3.758*** 4.408*** 4.420*** 4.451*** (0.632) (0.659) (0.579) (.577) (0.563) HAR 3.824** 2.706 2.757 (1.799) (2.207) (1.899) CII·HAR –2.621** –3.053** –2.829** (1.082) (1.190) (1.228) D_Prov_capital –0.765 –0.761 –1.893 –1.954 (1.463) (1.484) (2.215) (2.285) R2 0.733 0.735 0.752 0.754 0.752 Notes: Robust standard errors in parentheses; *** Significant at 1%; ** Significant at 5%; * Significant at 10%.

As shown in specification (5), we also dropped the reconstruction fund and the institutional variable in the estimation since both variables are consistently insignificant. Also the results of this estimation reveal a robust impact of other dependent variables on post-disaster growth. Therefore, it confirms temporary impact of disasters and the positive impact of peace deal on the population growth in Aceh and North Sumatera provinces after affected by the tsunami and earthquake.

5. Conclusion In this paper we analyze the impact of the tsunami in 2004 and the earthquake in 2005 on population dynamic across administrative units of the provinces of Aceh and North Sumatera building on the approach developed by Davis and Weinstein (2002). The results show that

21 there is no statistically significant change of the relative position of regions in this province although they have been shocked by huge exogenous disasters. The twin disasters have negatively affected the relative position of some highly affected regions, but most of them have completely recovered their rank and size in terms of population. This finding strongly points at the relevance of the locational fundamental theory, rather than the random growth theory. Furthermore, we also find that the twin disasters did not affect the regularity of size- distribution of regions in Aceh and North Sumatera. Zipf’ law exists even in the group of regions affected by a giant tsunami and a big earthquake. Estimations on post-disaster population growth provide strong evidence that the twin disasters had only a temporary impact. In other words, the results tend to reject the random growth theory, but support locational fundamental theory. This evidence is further supported by the impact of conflict intensity index that indicates that regions that suffered from past conflict have benefited from the peace deal.

References Andersen, G. and Y. Ge (2005), The Size Distribution of Chinese Cities, Regional Science and Urban Economics, 35, 756–776. Barro, R.J. (2006), Rare Disasters and Asset Markets in the Twentieth Century, Quarterly Journal of Economics, 121, 823–866. Barro, R.J. (2009), Rare Disasters, Asset Prices, and Welfare Costs, American Economic Review, 99, 243–264. Berry, B.J.L. and A. Okulicz-Kozaryn (2011), The City Size Distribution Debate: Resolution for US Urban Regions and Megalopolitan Areas, Cities, 29, S17–S23. Billon, P.L. and A. Waizenegger (2007), Peace in the Wake of Disaster? Secessionist Conflicts and the 2004 Indian Ocean Tsunami, Transactions of the Institute of British Geographers, 32, 411–427. Bosker, M., S. Brakman, J.H. Garretsen and M. Schramm (2008), A Century of Shocks: The Evolution of the German City Size Distribution 1925–1999, Regional Science and Urban Economics, 38, 330–347.

22

Badan Pusat Statistik (2010), Migrasi Internal Penduduk Indonesia Hasil Sensus Penduduk 2010, Jakarta, BPS. Badan Pusat Statistik (2010), Government of Aceh, United Nations Development Programme 2010, Provincial Human Development Report Aceh 2010, Jakarta, UNDP. Brakman, S., J.H. Garretsen and M. Schramm (2004), The Strategic Bombing of German Cities during World War II and its Impact on City Growth, Journal of Economic Geography, 4, 201–218. Chen, Z., S. Fu and D. Zhang (2010), Searching for the Parallel Growth of Cities (retrieved from http://mpra.ub.uni-muenchen.de/21528/1/MPRA_paper_21528.pdf on April 30, 2012). Cavallo, E. and I. Noy (2011), Natural Disasters and the Economy – A Survey, International Review of Environmental and Resource Economics, 5, 63–102. Cavallo, E., S. Galiani, I. Noy and J. Pantano (2011), Catastrophic Natural Disasters and Economic Growth (retrieved from http://www.artsci.wustl.edu/~pantano/Disasters.pdf on June 10, 2012). Davis, D.R. and D.E. Weinstein (2002), Bones, Bombs and Break Points: The Geography of Economic Activity, American Economic Review, 92, 1269–1289. Davis, D.R. and D.E. Weinstein (2008), A Search for Multiple Equilibria in Urban Industrial Structure, Journal of Regional Science, 48, 29–65. Gabaix, X. (2008), Variable Rare Disasters: An Exactly Solved Framework for Ten Puzzles in Macro-Finance, NBER Working Paper, no. 13724, Cambridge, MA. Gabaix, X. and Y.M. Ioannides (2004), The Evolution of City Size Distributions, in: J.V. Henderson and J.F. Thisse (ed.), Handbook of Regional and Urban Economics, volume 4, 2341–2378, Elsevier, North-Holland. Gabaix, X. and R. Ibragimov (2009), Rank – ½: A Simple Way to Improve OLS Estimation of Tail Exponents (retrieved from http:// SSRN-id881759.pdf on April 30, 2012). Gayatri, I.H. (2009), Dynamics of Village Institution: the Gampong under Aceh’s Special Autonomy, Journal of Indonesian Social Sciences and Humanities, 2, 203–212. Giesen, K. and J. Sudekum (2011), Zipf’s Law for Cities in the Regions and the Country, Journal of Economic Geography, 11, 667–686.

23

González-Val, R. and M. Sanso-Navarro (2010), Gibrat’s Law for Countries, Journal of Population Economics, 23, 1371–1389. Fitrani, F., B. Hofman and K. Kaiser (2005), Unity in Diversity? The Creation of New Local Governments in a Decentralising Indonesia, Bulletin of Indonesian Economic Studies, 41, 57–79. Fujita, M. and T. Mori (1996), The Role of Ports in the Making of Major Cities: Self- Agglomeration and Hub-Effect, Journal of Development Economics, 49, 93–120. Kennedy, J., J. Ashmore, E. Babister and I. Kelman (2008), The Meaning of ‘Build Back Better’: Evidence from Post-Tsunami Aceh and Sri Lanka, Journal of Contingencies and Crisis Management, 16, 24–36. Krugman, P. (1996), Confronting the Mystery of Urban Hierarchy, Journal of the Japanese and International Economies, 10, 399–418. Mahdi, S. (2007), Where do IDPs Go? Evidence of Social Capital from Aceh Conflict and Tsunami IDPs (retrieved from http://www.conflictrecovery.org/bin/Saiful_Mahdi-Where_do_IDPs_go.pdf on July 7, 2012). Masyrafah, H. and J.M.J.A. McKeon (2008), Post-tsunami Aid Effectiveness in Aceh, Proliferation and Coordination in Reconstruction, Wolfenshohn Center for Development Working Paper no. 6, The Brookings Institute, Washington. Miguel, E. and G. Roland (2011), The Long-run Impact of Bombing Vietnam, Journal of Development Economics, 96, 1–15. Naik, A., E. Stigter and F. Laczko (2007), Migration, Development and Natural Disasters: Insights from the Indian Ocean Tsunami, IOM Migration Research Series, no. 3, International Organization for Migration, Geneva. Nazara, S. and B.P. Resosudarmo (2007), Aceh-Nias Reconstruction and Rehabilitation: Progress and Challenges at the End of 2006, ADB Institute, Tokyo. Nitsch, V. (2005), Zipf Zipped, Journal of Urban Economics, 57, 86–100. Paul, B.K. (2005), Evidence against Disaster-induced Migration: The 2004 Tornado in North- central Bangladesh, Disasters, 29, 370–385. Polèse , M. and J. Denis-Jacob (2010), Changes at the Top: A Cross-country Examination over the 20th Century of Rise (and Fall) in Rank of the Top Cities in National Urban Hierarchies, Urban Studies, 47, 1843–1860.

24

Takahashi, M., S. Tanaka, R. Kimura, R. Tabuchi, T. Kuroda, M. Ando and F. Kimata (2007), Restoration after the Sumatra Tsunami in Banda Aceh: Based on the Results of Interdisciplinary Researches by Nagoya University, Journal of Natural Disaster Science, 29, 53–61. Soo, K.T. (2007), Zipf’s Law and Urban Growth in Malaysia, Urban Studies, 44, 1–14. Soo, K.T. (2010), Zipf, Gibrat and Geography: Evidence from China, India and Brazil, (retrieved from http://www.lancs.ac.uk/staff/sook/papers/CIB%203%20web.pdf on April 29, 2012). Vale, L.J. and T.J. Campanella (2005), Conclusion-Axioms of Resilience, in L.J. Vale and T.J. Campanella (eds), The Resilience City: How Modern Cities Recover from Disaster, Oxford University Press, New York. World Bank (2009), Aceh Growth Diagnostic: Identifying the binding constraints to growth in a post-conflict and post-disaster environment, (retrieved on 28 April 2012 from http://www.hks.harvard.edu/fs/drodrik/Growth%20diagnostics%20papers/Aceh%20Grow th%20Diagnostic_english.pdf).

25

Appendix 1. Abbreviations of names of regions

No Region Abbvrev No Region Abbvrev 1 Aceh Barat ABr 21 Nagan Raya NR 2 Aceh Barat Daya ABD 22 Nias Nias 3 Aceh Besar Abs 23 Pidie Pid 4 Aceh Jaya AJ 24 Simalungun Sim 5 Aceh Selatan AS 25 Simeulue Sima 6 Aceh Singkil Asi 26 TapanuliSelatan TS 7 Aceh Tamiang Atam 27 TapanuliTengah Tteng 8 Aceh Tengah Ateng 28 TapanuliUtara TU 9 Aceh Tenggara AT 29 TobaSamosir Tsa 10 Aceh Timur Atim 30 Banda Aceh BA 11 Aceh Utara AU 31 Bin 12 Asahan Asah 32 Langsa Lang 13 Bireuen Bire 33 Lhokseumawe Lhok 14 Dairi Dai 34 Medan Med 15 DeliSerdang DS 35 Psia 16 Gayo Lues GL 36 Sabang Sab 17 K a r o Karo 37 Sib 18 LabuhanBatu LB 38 TjB 19 Langkat Langkt 39 TebingTinggi TT 20 MandailingNatal MN

Appendix 2. Data sources11

Population data are taken from the BPS-Statistics Indonesia. Aceh and Nias population data for 2005 is based on the Population Census in Aceh and Nias (SPAN 2005). The BPS conducted this census by using a “Moment Telling” method in August and September 2005 with supports from UNFPA, CIDA, AusAID, and NZAid (UNFPA 2006). Visitors intending to stay for less than six months were not enumerated. The main aim of this census is to provide data that were important to plan reconstruction programs. This is said, for instance by the UNFPA, as the quickest and the most complete population count especially in Aceh since

11 For regions that were not affected by disasters, especially most regions in North Sumatera, we assumed the related-disaster variables to be zero. The same approach is applied for conflict data since the survey for this data did not cover some regions in Aceh (mainly cities) as well as all regions in North Sumatera.

26 there was a security issue in this area in the previous census. Meanwhile, the source of data for 2010 is the Population Census 2010 conducted by the BPS.

Data on Death and missing people on Aceh are taken from the Indonesian Disaster Information and Data (DIBI) unit of the National Disaster Management Agency (DIBI- BNPB).12 Data for Nias are taken from BRR NAD-Nias Perwakilan Nias 2008, Laporan Akhir BRR NAD-Nias Perwakilan Nias 2005–2008, pp. 5–6 (retrieved on May 1, 2012 from http://monevacehnias.bappenas.go.id/images/kumpulan_laporan/laporan_brr/BUKU_LAPOR AN_AKHIR_NIAS.pdf).

Internally displaced people (IDP) in Aceh is taken from BRR, UN, Pemprov NAD, 2009, Tsunami Recovery Indicator Package (TRIP), the third report for Aceh and Nias, p. 66, (retrieved from http://monevacehnias.bappenas.go.id/images/kumpulan_laporan/ laporan_donor_lsm/TRIP_2009_Report.pdf on May 2, 2012). The Nias data are taken from Laporan Akhir BRR Aceh-Nias Perwakilan Nias 2005–2008, p. 6 (retrieved on May 1, 2012 from http://monevacehnias.bappenas.go.id/images/kumpulan_laporan/laporan_brr/ BUKU_LAPORAN_AKHIR_NIAS.pdf).

Houses damaged data are taken from BRR, UN, Pemprov NAD, 2009, Tsunami Recovery Indicator Package (TRIP), the third report for Aceh and Nias, January 2009, p. 66 (retrieved from http://monevacehnias.bappenas.go.id/images/kumpulan_laporan/laporan_donor_lsm/ TRIP_2009_Report.pdf on May 2, 2012).

The Conflict intensity index is derived from Kecamatan Development Program, Ministry of Home Affairs, World Bank (2007), 2006 Village Survey in Aceh, An Assessment of Village Infrastructure and Social Conditions, p. 36 (retrieved on April 29, 2012, from

12 Data can be accessed at http://dibi.bnpb.go.id. BNPB was established in 2008 under the Regulation of the President No. 8/2008 and the objective of DIBI is to provide data for risk identification, policy formulation and decision making. The methodologies used by DIBI are DesInventar and Desconsular (http://www.desinventar.org/en/). Some reports on Aceh and Nias provide data on disaster impact, but we chose to use the DIBI’s data since it covers Indonesia as a whole so that it provides flexibility if we need to extend our sample. Besides, for several regions, the BRR’s reports do not provide data on death toll and missing people.

27 http://siteresources.worldbank.org/INTINDONESIA/Resources/226271-1168333550999/ AcehVillageSurvey06_final.pdf).

Disbursed funds are based on Recovery Aceh Nias Database (RAND) (retrieved on May 23, 2012 from http://rand.bappenas.go.id/RAND/rc?sessionid=1336654759391180). These data do not include data for “All Kabupaten–All Provinces”, “Unallocated”, “Kabupaten to be defined – All Provinces”, and “Unspecified”. RAND provides data in US dollar. It covers projects starting from 6 October 2005 to 31 July 2009 and the last modified is on 28 September 2010 (see http://rand.bappenas.go.id/RAND/rc?requesttype=entry&entryid= ReportEntry&reportid=2042&UID=1_12_0_48_48_0).

28